How to write good test suites for Ren'Py games?

[SoGu]

I'm currently writing a framework for creating Ren'Py games based on character personality models. This framework consists of several small Python scripts (and is still supposed to grow). I really want to test everything properly, because the framework's complexity is quite ugly for such a small script.

However, I don't know how to tackle this testing. Most important for me would be unit tests. Probably making a test "game" with a start label and letting lots of small tests run would be possible but time-consuming. Is there an elegant way of automating test execution? Fast-forwarding could work well, I guess, but only if there is an easy way of interrupting fast-forwarding when a test case fails.

Has anyone else had a similar problem or an idea how unit-testing within Ren'Py could be solved in the most elegant way?

[PyTom]

One way to do this would be to isolate out the python code, and test it separately from the Ren'Py interface. There's also a Ren'Py testcase language, although it's more for testing gui code than logic - you can see it in testcases.rpy in the launcher/game and tutorial/game directories of recent Ren'Pys.

[Apa]

Has anyone else had a similar problem or an idea how unit-testing within Ren'Py could be solved in the most elegant way?

Could you give a few examples of such tests (with as success and failure outcomes), please?

The only test I can think about is “game logic test” at very high level:
Get all labels, make graph and find unreachable [groups of] nodes (aka garbage collector test)
Tho, it’s not any easy to do because of jumps/calls to dynamic (calculated at run-time) labels, so some sort of hints might be needed.

Another idea is an "automated game walker" to traverse all/chosen path from start/save point, so you don’t have to click anything, but only watch (to fix a bug at specific game stage). Not sure how usable it’ll be either.

I don’t have any more ideas as of now...